The wire-level streaming vocabulary of [dsh-llm](../../packages/llm/llm). [core.md](core.md) introduces `StreamChunk`, `Message`, and `ContentBlock`; this page owns the full chunk protocol, the adapter contract every adapter must obey, and the shared assembler.
A streaming response interleaves several typed blocks (text, reasoning, multiple tool calls). `index` ties each delta to its block; `block-end` carries the fully-assembled `ContentBlock` so consumers don't have to re-assemble deltas themselves. It is a **closed** discriminated union — a `switch` over `type` ends with `assertNever`, so adding a variant breaks compilation at every consumer that must handle it.
Every thrown or in-band final-adapter failure normalizes to one serializable provider-neutral payload. `providerRetryAfterMs` is a validated positive delay requested by the provider, not a retry decision; `ProviderRequestId` is an opaque branded string for diagnostics.
Every adapter MUST obey these, and every consumer may rely on them:
- **`usage` before `finish`, nothing after `finish`.** Defer both to the provider's end-of-stream marker so a trailing usage-only chunk can't violate the ordering.
- **Tool-call `arguments` stay raw JSON strings end-to-end.** Partial fragments stream via `argumentsDelta`; a provider that hands back parsed objects re-stringifies at `block-end`.
- **Two sanctioned error paths, one fact shape.** A failure may either THROW from `stream()` (transport/protocol errors) **or** end the stream with `finish {kind:'error'|'aborted', failure}` (provider in-band errors, for adapters that can't throw mid-stream). `LlmError.failure` carries the same `LlmFailure`. The final adapter boundary preserves the exact thrown `Error` object and associates immutable facts plus the serving registration's immutable retry policy with that call; the agent loop closes the failed step and offers the error, facts, immutable prior-retried facts, serving policy, and turn signal to `agent/request-error`. A handling listener returns `{ kind: 'retry' }` after its awaited repair; absent recovery the structured failure becomes the turn error, and no normal assistant message or tool side effect is committed for that attempt.
- **One adapter call is one provider attempt.** Adapters disable library retries. Agent-level recovery opens another durable numbered turn; direct `ctx.llm.stream()` callers remain single-attempt.
- **Provider stalls are bounded at the transport.** Both shipping remote adapters expose positive finite `streamIdleTimeoutMs` with a five-minute default. The watchdog arms only while iterator `next()` is outstanding, uses one stable signal for the whole request, maps its own expiry to `TIMEOUT`, and keeps an earlier caller abort as `ABORTED`.
- **Context overflow has one canonical code.** Both DeepSeek adapters classify explicit provider detail through `isContextWindowExceededError()` and surface `CONTEXT_WINDOW_EXCEEDED`, whether the failure arrives as a thrown HTTP `LlmError` or an in-band finish error. Consumers route on the code, never provider text.
- **An empty completion is a retryable error, not a silent success.** Both adapters map a terminal `stop` finish that carried no content blocks to `finish {kind:'error'}` with the canonical `EMPTY_RESPONSE` code, and `dsh-llm-retry` retries it by default; see [empty model responses are retryable](../../.agents/notes/implemented/bug-fix/2026-07-24-empty-model-response-is-retryable.md).
- **Replay state is adapter-owned.** A successful `finish` may carry lossless-JSON state needed to reconstruct a native provider response. The loop stores it with the assembled assistant message. On a later request, `LlmService` passes the state only when the historical provider and target provider are currently registered to the exact same adapter instance. That adapter validates the state and owns any cross-model or cross-provider conversion; other adapters receive the provider-neutral content and provenance without the private state.
Two independent implementations obey this contract: `dsh-llm-deepseek` uses direct fetch with SSE framing through `eventsource-parser`, while `dsh-llm-pi-ai` provides a generic multi-provider adapter through `@earendil-works/pi-ai`. Both carry cancellation and the idle watchdog to the provider request.
Provider configuration resolves before route registration into an immutable discriminated union. Normal mode carries `mode: 'normal'`, finite `maxRetries`, `retryableCodes`, and required `initialDelayMs`, `maxDelayMs`, and `jitterRatio`; always mode carries `mode: 'always'` and the same required backoff fields without a finite maximum. `LlmService.providerRetryPolicy(provider)` returns the currently registered value and supplies normal defaults when the adapter omits one; `llmRetryPolicyOf(stream)` returns the exact serving registration's captured value after that call enters its final adapter boundary, so later route disposal or replacement cannot change an in-flight failure's recovery policy. The [generated config catalog](../config-catalog.md) owns the optional input shapes.
The static public application identity every adapter sends to providers ([`packages/llm/llm/src/attribution.ts`](../../packages/llm/llm/src/attribution.ts)). `attributionHeaders(identity?)` maps it to the standard `User-Agent` header only; OpenRouter-specific app attribution headers are intentionally not supported by this contract. The default `APP_IDENTITY` sources its version from the package manifest; every field is a public product fact - no secrets, paths, session ids, or per-user identifiers, and nothing per-request may influence the values. Rationale: [Mandatory `User-Agent` attribution](../../.agents/notes/implemented/architecture/2026-06-21-mandatory-app-attribution-headers.md).
Per-call token accounting. Counts are **disjoint**: `inputTokens` is uncached input only; cached input is reported separately, and billed input is the sum of the three. Adapters whose providers fold cache hits into a single prompt total (DeepSeek's `prompt_tokens`) subtract them back out. `reasoningTokens`, when present, is informational detail already included in `outputTokens`; totals must not add it again.
`BlockAssembler` ([`packages/llm/llm/src/assembler.ts`](../../packages/llm/llm/src/assembler.ts)) is the single shared implementation that folds a `StreamChunk` stream back into `ContentBlock`s, usage, finish reason, and replay state. The loop logs the raw chunks while feeding the same chunks through an assembler, then stores the assembled assistant content with its provider/model provenance. A consumer that needs the assembled result without re-implementing the fold uses this.
`LlmAdapter` is the provider seam: subclass, implement `stream()`, and register one adapter instance with `ctx.llm.registerAdapter(providers, adapter)`. `GenerateOptions.provider` selects the registered adapter; `GenerateOptions.model` is passed to that adapter and need not be registered at lifecycle start. Duplicate provider routes fail atomically. Optional `providerRetryPolicy()` is captured per route with normal defaults, while `providerInfo()` and asynchronous `listModels()` feed `LlmService.listProviders()` / `listModels()` with detached selector metadata. That catalog is advisory rather than a request whitelist: the adapter remains authoritative and may accept unlisted model ids. One asynchronous `resolveModel()` query returns exact model identity plus optional correctness-sensitive context capacity, an adapter-configured `defaultMaxTokens`, and ordered model-owned reasoning ids with an optional deployment default; absent fields mean unavailable metadata or provider-owned behavior, not invalid catalog membership. The resolver receives optional cancellation and must settle promptly after abort. `LlmService.resolveModelInfo()` validates and detaches the aggregate. At the final adapter boundary, `resolveCallConfig()` materializes the output default only when `maxTokens` is absent and validates and materializes reasoning, so direct calls cannot bypass either configured behavior; direct dispatch captures one registration before awaiting that resolution. The agent loop instead uses `prepareCall()` to keep the same registration across model resolution, durable header logging, and dispatch, retain detached context metadata from that exact lookup, and report which config fields the adapter defaulted. Adapter lookup happens at the terminal continuation of the `llm/stream` waterfall, so a listener may short-circuit the call or route a mutable one-shot request before lookup. AgentLoop observes a request attempt once the outer waterfall returns a stream handle; that limited boundary does not prove a lazy terminal adapter was constructed or began provider I/O. The `block-start` / `block-end` `index` correlation and the assembler together mean an adapter only has to emit well-formed chunks — block reassembly is not each adapter's problem. The consumer surface (`ctx.llm.stream()`) and the `llm/stream` waterfall are described in [architecture.md § Content blocks and streaming](../architecture.md#content-blocks-and-streaming-dsh-llm).
